Project Summary Myocardial first-pass perfusion and late gadolinium enhancement (LGE) schemes are key components of most clinical cardiac MRI exams. The limitations of current MRI schemes often makes it challenging to simultaneously achieve high spatio-temporal resolution, sufficient spatial coverage, and good image quality in first-pass perfusion MRI, making it difficult to interpreting the results. Similarly, the large number of breath-holds and their long duration often makes LGE acquisitions challenging for many patients, resulting in significant motion artifacts and reduced patient throughput. In this context, there is an immediate clinical need for a novel dynamic imaging framework that can enable free-breathing acquisitions and considerably improve spatio-temporal resolution and coverage, without degrading the quality. The main objective of this proposal is to develop a novel dynamic imaging framework, which can enable free-breathing cardiac MRI and significantly accelerate it with minimal artifacts. We recently introduced a novel regularized reconstruction algorithm to significantly accelerate free-breathing dynamic MRI data. Preliminary validations of the algorithm demonstrated the ability of the proposed scheme to provide accelerations of up-to eleven fold with minor artifacts. The main focus of this proposal is to further improve the k-t SLR scheme and use it to realize high-resolution clinical myocardial perfusion and free-breathing LGE MRI. The successful completion of the proposed research will provide quantitative perfusion estimates with a temporal resolution of one heartbeat and spatial resolution of 0.15x0.15x0.8 ccfrom the entire heart, which is a four-fold improvement over current schemes. Similarly, we expect to considerably improve the patient compliance by relaxing the breath-holding requirement and reducing the scan time in LGE MRI data. These developments are quite significant and will considerably advance the state of the art in contrast-enhanced CMRI. The proposed algorithm is a radical departure from the classical approaches that rely on x-f space sparsity. In addition, we introduce non-convex spectral priors and additionally exploit the sparsity of the dynamic images to further improve the data fidelity and acceleration rate. Thus, the proposed scheme is highly innovative and its impact is expected to extend beyond the specific applications. Our team is well qualified to perform the proposed research because of our combined scope and breadth in expertise (including signal/image processing, MR physics, radiology, and cardiology), in addition to the extensive preliminary data.